Am Freitag, den 28.02.2014, 13:27 +0100 schrieb Arnd Bergmann:

On Friday 28 February 2014 13:09:34 Lucas Stach wrote:

quoted

Am Freitag, den 28.02.2014, 12:37 +0100 schrieb Arnd Bergmann:

quoted

On Friday 28 February 2014 12:27:36 Lucas Stach wrote:
The interrupt-map doesn't even look well-formed, since it refers to
a gic that has #interrupt-cells=<3>, but only has one cell with the
value. This clearly cannot work at all. It should probably be

      interrupt-map = <0x0 0 &gic GIC_SPI 21 IRQ_TYPE_LEVEL_HIGH>;

Note that '53' is the IRQ number that is commonly assigned to
SPI interrupt 21, since SPI starts at number 32 on GIC.

Ah, I see. Thanks for the explanation. So we have a bogus Exynos DT,
which will break existing systems once the kernel starts to do the right
thing. This is starting to look great. 

 
It seems tegra and r-car have the same bug. We can probably fix it
in a backwards-compatible way by changing the function to call
the generic helper first and if that fails call printk_once()
to emit a warning and return the hardwired IRQ number.

Yes, seems like the way to go. I'll prepare some patches for that. It
seems of_irq_parse_and_map_pci() already emits an error message if it
can't find a valid mapping. This should be enough to get people to fix
their DTs.

Regards,
Lucas 

-- 
Pengutronix e.K.                           | Lucas Stach                 |
Industrial Linux Solutions                 | http://www.pengutronix.de/  |
Peiner Str. 6-8, 31137 Hildesheim, Germany | Phone: +49-5121-206917-5076 |
Amtsgericht Hildesheim, HRA 2686           | Fax:   +49-5121-206917-5555 |